End of training

README.md

@@ -1,21 +1,24 @@
 ---
+language:
+- gl
 license: apache-2.0
 base_model: openai/whisper-tiny
 tags:
+- whisper-event
 - generated_from_trainer
 datasets:
-- common_voice_13_0
+- mozilla-foundation/common_voice_13_0
 metrics:
 - wer
 model-index:
-- name: openai/whisper-tiny
+- name: Whisper Tiny Galician
   results:
   - task:
       name: Automatic Speech Recognition
       type: automatic-speech-recognition
     dataset:
-      name: common_voice_13_0
-      type: common_voice_13_0
+      name: mozilla-foundation/common_voice_13_0 gl
+      type: mozilla-foundation/common_voice_13_0
       config: gl
       split: test
       args: gl
@@ -28,9 +31,9 @@ model-index:
 <!-- This model card has been generated automatically according to the information the Trainer had access to. You
 should probably proofread and complete it, then remove this comment. -->
 
-# openai/whisper-tiny
+# Whisper Tiny Galician
 
-This model is a fine-tuned version of [openai/whisper-tiny](https://huggingface.co/openai/whisper-tiny) on the common_voice_13_0 dataset.
+This model is a fine-tuned version of [openai/whisper-tiny](https://huggingface.co/openai/whisper-tiny) on the mozilla-foundation/common_voice_13_0 gl dataset.
 It achieves the following results on the evaluation set:
 - Loss: 0.6003
 - Wer: 26.1331

all_results.json

@@ -1,12 +1,12 @@
 {
-    "epoch": 99.01,
-    "eval_loss": 0.5832294821739197,
-    "eval_runtime": 845.9676,
-    "eval_samples_per_second": 7.738,
-    "eval_steps_per_second": 0.061,
-    "eval_wer": 26.35037251655629,
-    "train_loss": 0.05570581116452813,
-    "train_runtime": 137346.4264,
-    "train_samples_per_second": 9.319,
-    "train_steps_per_second": 0.036
+    "epoch": 100.0,
+    "eval_loss": 0.6002562642097473,
+    "eval_runtime": 42.818,
+    "eval_samples_per_second": 152.88,
+    "eval_steps_per_second": 1.214,
+    "eval_wer": 26.13307119205298,
+    "train_loss": 0.5713151776909828,
+    "train_runtime": 5322.8653,
+    "train_samples_per_second": 240.472,
+    "train_steps_per_second": 0.939
 }

eval_results.json

@@ -1,8 +1,8 @@
 {
-    "epoch": 99.01,
-    "eval_loss": 0.5832294821739197,
-    "eval_runtime": 845.9676,
-    "eval_samples_per_second": 7.738,
-    "eval_steps_per_second": 0.061,
-    "eval_wer": 26.35037251655629
+    "epoch": 100.0,
+    "eval_loss": 0.6002562642097473,
+    "eval_runtime": 42.818,
+    "eval_samples_per_second": 152.88,
+    "eval_steps_per_second": 1.214,
+    "eval_wer": 26.13307119205298
 }

runs/Jan02_16-03-37_hyperion-255/events.out.tfevents.1735835766.hyperion-255

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:20dc660e88429f172d44bc31dd454e7f6f4318f580161013f2e09c0561d68e8b
+size 40

train_results.json

@@ -1,7 +1,7 @@
 {
-    "epoch": 99.01,
-    "train_loss": 0.05570581116452813,
-    "train_runtime": 137346.4264,
-    "train_samples_per_second": 9.319,
-    "train_steps_per_second": 0.036
+    "epoch": 100.0,
+    "train_loss": 0.5713151776909828,
+    "train_runtime": 5322.8653,
+    "train_samples_per_second": 240.472,
+    "train_steps_per_second": 0.939
 }

trainer_state.json

@@ -1,7 +1,7 @@
 {
-  "best_metric": 26.35037251655629,
-  "best_model_checkpoint": "./checkpoint-1000",
-  "epoch": 99.01,
+  "best_metric": 26.13307119205298,
+  "best_model_checkpoint": "./checkpoint-5000",
+  "epoch": 100.0,
   "eval_steps": 1000,
   "global_step": 5000,
   "is_hyper_param_search": false,
@@ -9,1265 +9,1267 @@
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.01,
+      "epoch": 0.5,
       "learning_rate": 1.575e-06,
-      "loss": 2.2585,
+      "loss": 10.212,
       "step": 25
     },
     {
-      "epoch": 0.01,
+      "epoch": 1.0,
       "learning_rate": 3.4499999999999996e-06,
-      "loss": 1.6463,
+      "loss": 8.6948,
       "step": 50
     },
     {
-      "epoch": 1.0,
+      "epoch": 1.5,
       "learning_rate": 5.324999999999999e-06,
-      "loss": 1.0949,
+      "loss": 7.0414,
       "step": 75
     },
     {
-      "epoch": 1.01,
+      "epoch": 2.0,
       "learning_rate": 7.2e-06,
-      "loss": 0.7923,
+      "loss": 6.0592,
       "step": 100
     },
     {
-      "epoch": 2.0,
+      "epoch": 2.5,
       "learning_rate": 9.074999999999999e-06,
-      "loss": 0.6811,
+      "loss": 5.4484,
       "step": 125
     },
     {
-      "epoch": 2.01,
+      "epoch": 3.0,
       "learning_rate": 1.0949999999999998e-05,
-      "loss": 0.5602,
+      "loss": 5.0744,
       "step": 150
     },
     {
-      "epoch": 3.0,
+      "epoch": 3.5,
       "learning_rate": 1.2825e-05,
-      "loss": 0.5154,
+      "loss": 4.8184,
       "step": 175
     },
     {
-      "epoch": 3.01,
+      "epoch": 4.0,
       "learning_rate": 1.47e-05,
-      "loss": 0.4336,
+      "loss": 4.623,
       "step": 200
     },
     {
-      "epoch": 4.0,
+      "epoch": 4.5,
       "learning_rate": 1.6575e-05,
-      "loss": 0.4093,
+      "loss": 4.4652,
       "step": 225
     },
     {
-      "epoch": 4.01,
+      "epoch": 5.0,
       "learning_rate": 1.8449999999999998e-05,
-      "loss": 0.3408,
+      "loss": 4.3383,
       "step": 250
     },
     {
-      "epoch": 5.0,
+      "epoch": 5.5,
       "learning_rate": 2.0325e-05,
-      "loss": 0.3255,
+      "loss": 4.2083,
       "step": 275
     },
     {
-      "epoch": 5.01,
+      "epoch": 6.0,
       "learning_rate": 2.2199999999999998e-05,
-      "loss": 0.2656,
+      "loss": 4.1224,
       "step": 300
     },
     {
-      "epoch": 6.0,
+      "epoch": 6.5,
       "learning_rate": 2.4075e-05,
-      "loss": 0.2539,
+      "loss": 4.001,
       "step": 325
     },
     {
-      "epoch": 6.01,
+      "epoch": 7.0,
       "learning_rate": 2.5949999999999997e-05,
-      "loss": 0.201,
+      "loss": 3.9214,
       "step": 350
     },
     {
-      "epoch": 7.0,
+      "epoch": 7.5,
       "learning_rate": 2.7825e-05,
-      "loss": 0.1936,
+      "loss": 3.6657,
       "step": 375
     },
     {
-      "epoch": 7.01,
+      "epoch": 8.0,
       "learning_rate": 2.97e-05,
-      "loss": 0.1458,
+      "loss": 2.9008,
       "step": 400
     },
     {
-      "epoch": 8.01,
+      "epoch": 8.5,
       "learning_rate": 3.1574999999999995e-05,
-      "loss": 0.1416,
+      "loss": 1.9626,
       "step": 425
     },
     {
-      "epoch": 8.01,
+      "epoch": 9.0,
       "learning_rate": 3.345e-05,
-      "loss": 0.1008,
+      "loss": 1.5518,
       "step": 450
     },
     {
-      "epoch": 9.01,
+      "epoch": 9.5,
       "learning_rate": 3.5325e-05,
-      "loss": 0.0983,
+      "loss": 1.2819,
       "step": 475
     },
     {
-      "epoch": 9.01,
+      "epoch": 10.0,
       "learning_rate": 3.7199999999999996e-05,
-      "loss": 0.0675,
+      "loss": 1.1488,
       "step": 500
     },
     {
-      "epoch": 10.01,
+      "epoch": 10.5,
       "learning_rate": 3.732499999999999e-05,
-      "loss": 0.0681,
+      "loss": 1.0035,
       "step": 525
     },
     {
-      "epoch": 10.01,
+      "epoch": 11.0,
       "learning_rate": 3.711666666666666e-05,
-      "loss": 0.0453,
+      "loss": 0.9323,
       "step": 550
     },
     {
-      "epoch": 11.01,
+      "epoch": 11.5,
       "learning_rate": 3.690833333333333e-05,
-      "loss": 0.0487,
+      "loss": 0.8265,
       "step": 575
     },
     {
-      "epoch": 11.01,
+      "epoch": 12.0,
       "learning_rate": 3.67e-05,
-      "loss": 0.0333,
+      "loss": 0.7946,
       "step": 600
     },
     {
-      "epoch": 12.01,
+      "epoch": 12.5,
       "learning_rate": 3.649166666666667e-05,
-      "loss": 0.0377,
+      "loss": 0.7179,
       "step": 625
     },
     {
-      "epoch": 12.01,
+      "epoch": 13.0,
       "learning_rate": 3.6283333333333325e-05,
-      "loss": 0.026,
+      "loss": 0.6862,
       "step": 650
     },
     {
-      "epoch": 13.01,
+      "epoch": 13.5,
       "learning_rate": 3.6074999999999996e-05,
-      "loss": 0.0277,
+      "loss": 0.6252,
       "step": 675
     },
     {
-      "epoch": 13.01,
+      "epoch": 14.0,
       "learning_rate": 3.586666666666666e-05,
-      "loss": 0.02,
+      "loss": 0.6159,
       "step": 700
     },
     {
-      "epoch": 14.01,
+      "epoch": 14.5,
       "learning_rate": 3.565833333333333e-05,
-      "loss": 0.0215,
+      "loss": 0.5589,
       "step": 725
     },
     {
-      "epoch": 14.01,
+      "epoch": 15.0,
       "learning_rate": 3.545e-05,
-      "loss": 0.0157,
+      "loss": 0.5586,
       "step": 750
     },
     {
-      "epoch": 15.01,
+      "epoch": 15.5,
       "learning_rate": 3.5241666666666665e-05,
-      "loss": 0.0167,
+      "loss": 0.5058,
       "step": 775
     },
     {
-      "epoch": 15.01,
+      "epoch": 16.0,
       "learning_rate": 3.503333333333333e-05,
-      "loss": 0.0132,
+      "loss": 0.5053,
       "step": 800
     },
     {
-      "epoch": 16.0,
+      "epoch": 16.5,
       "learning_rate": 3.482499999999999e-05,
-      "loss": 0.0138,
+      "loss": 0.4617,
       "step": 825
     },
     {
-      "epoch": 16.01,
+      "epoch": 17.0,
       "learning_rate": 3.461666666666666e-05,
-      "loss": 0.0108,
+      "loss": 0.4611,
       "step": 850
     },
     {
-      "epoch": 17.0,
+      "epoch": 17.5,
       "learning_rate": 3.4408333333333334e-05,
-      "loss": 0.0113,
+      "loss": 0.4201,
       "step": 875
     },
     {
-      "epoch": 17.01,
+      "epoch": 18.0,
       "learning_rate": 3.42e-05,
-      "loss": 0.0088,
+      "loss": 0.4233,
       "step": 900
     },
     {
-      "epoch": 18.0,
+      "epoch": 18.5,
       "learning_rate": 3.399166666666667e-05,
-      "loss": 0.0093,
+      "loss": 0.3921,
       "step": 925
     },
     {
-      "epoch": 18.01,
+      "epoch": 19.0,
       "learning_rate": 3.3783333333333326e-05,
-      "loss": 0.0079,
+      "loss": 0.3853,
       "step": 950
     },
     {
-      "epoch": 19.0,
+      "epoch": 19.5,
       "learning_rate": 3.3574999999999996e-05,
-      "loss": 0.0077,
+      "loss": 0.3585,
       "step": 975
     },
     {
-      "epoch": 19.01,
+      "epoch": 20.0,
       "learning_rate": 3.336666666666667e-05,
-      "loss": 0.0062,
+      "loss": 0.3626,
       "step": 1000
     },
     {
-      "epoch": 19.01,
-      "eval_loss": 0.5832294821739197,
-      "eval_runtime": 629.2972,
-      "eval_samples_per_second": 10.402,
-      "eval_steps_per_second": 0.083,
-      "eval_wer": 26.35037251655629,
+      "epoch": 20.0,
+      "eval_loss": 0.5407164096832275,
+      "eval_runtime": 47.2578,
+      "eval_samples_per_second": 138.517,
+      "eval_steps_per_second": 1.1,
+      "eval_wer": 30.846440397350992,
       "step": 1000
     },
     {
-      "epoch": 20.0,
+      "epoch": 20.5,
       "learning_rate": 3.315833333333333e-05,
-      "loss": 0.0059,
+      "loss": 0.3313,
       "step": 1025
     },
     {
-      "epoch": 20.01,
+      "epoch": 21.0,
       "learning_rate": 3.295e-05,
-      "loss": 0.005,
+      "loss": 0.3363,
       "step": 1050
     },
     {
-      "epoch": 21.0,
+      "epoch": 21.5,
       "learning_rate": 3.2741666666666665e-05,
-      "loss": 0.0047,
+      "loss": 0.3064,
       "step": 1075
     },
     {
-      "epoch": 21.01,
+      "epoch": 22.0,
       "learning_rate": 3.253333333333333e-05,
-      "loss": 0.0041,
+      "loss": 0.3133,
       "step": 1100
     },
     {
-      "epoch": 22.0,
+      "epoch": 22.5,
       "learning_rate": 3.2325e-05,
-      "loss": 0.004,
+      "loss": 0.2886,
       "step": 1125
     },
     {
-      "epoch": 22.01,
+      "epoch": 23.0,
       "learning_rate": 3.2116666666666664e-05,
-      "loss": 0.0035,
+      "loss": 0.2927,
       "step": 1150
     },
     {
-      "epoch": 23.0,
+      "epoch": 23.5,
       "learning_rate": 3.1908333333333334e-05,
-      "loss": 0.0036,
+      "loss": 0.268,
       "step": 1175
     },
     {
-      "epoch": 23.01,
+      "epoch": 24.0,
       "learning_rate": 3.17e-05,
-      "loss": 0.0036,
+      "loss": 0.2773,
       "step": 1200
     },
     {
-      "epoch": 24.0,
+      "epoch": 24.5,
       "learning_rate": 3.149166666666666e-05,
-      "loss": 0.0038,
+      "loss": 0.2511,
       "step": 1225
     },
     {
-      "epoch": 24.01,
+      "epoch": 25.0,
       "learning_rate": 3.128333333333333e-05,
-      "loss": 0.0034,
+      "loss": 0.2599,
       "step": 1250
     },
     {
-      "epoch": 25.0,
+      "epoch": 25.5,
       "learning_rate": 3.1074999999999996e-05,
-      "loss": 0.0033,
+      "loss": 0.2354,
       "step": 1275
     },
     {
-      "epoch": 25.01,
+      "epoch": 26.0,
       "learning_rate": 3.086666666666667e-05,
-      "loss": 0.003,
+      "loss": 0.2438,
       "step": 1300
     },
     {
-      "epoch": 26.0,
+      "epoch": 26.5,
       "learning_rate": 3.065833333333333e-05,
-      "loss": 0.0032,
+      "loss": 0.2191,
       "step": 1325
     },
     {
-      "epoch": 26.01,
+      "epoch": 27.0,
       "learning_rate": 3.0449999999999998e-05,
-      "loss": 0.0026,
+      "loss": 0.2311,
       "step": 1350
     },
     {
-      "epoch": 27.0,
+      "epoch": 27.5,
       "learning_rate": 3.0241666666666662e-05,
-      "loss": 0.0026,
+      "loss": 0.2125,
       "step": 1375
     },
     {
-      "epoch": 27.01,
+      "epoch": 28.0,
       "learning_rate": 3.003333333333333e-05,
-      "loss": 0.0023,
+      "loss": 0.2093,
       "step": 1400
     },
     {
-      "epoch": 28.0,
+      "epoch": 28.5,
       "learning_rate": 2.9824999999999997e-05,
-      "loss": 0.0024,
+      "loss": 0.1959,
       "step": 1425
     },
     {
-      "epoch": 28.01,
+      "epoch": 29.0,
       "learning_rate": 2.9616666666666664e-05,
-      "loss": 0.0021,
+      "loss": 0.2009,
       "step": 1450
     },
     {
-      "epoch": 29.0,
+      "epoch": 29.5,
       "learning_rate": 2.940833333333333e-05,
-      "loss": 0.0022,
+      "loss": 0.1843,
       "step": 1475
     },
     {
-      "epoch": 29.01,
+      "epoch": 30.0,
       "learning_rate": 2.9199999999999995e-05,
-      "loss": 0.002,
+      "loss": 0.1921,
       "step": 1500
     },
     {
-      "epoch": 30.0,
+      "epoch": 30.5,
       "learning_rate": 2.8991666666666662e-05,
-      "loss": 0.002,
+      "loss": 0.1717,
       "step": 1525
     },
     {
-      "epoch": 30.01,
+      "epoch": 31.0,
       "learning_rate": 2.878333333333333e-05,
-      "loss": 0.0019,
+      "loss": 0.1813,
       "step": 1550
     },
     {
-      "epoch": 31.0,
+      "epoch": 31.5,
       "learning_rate": 2.8574999999999997e-05,
-      "loss": 0.0019,
+      "loss": 0.1679,
       "step": 1575
     },
     {
-      "epoch": 31.01,
+      "epoch": 32.0,
       "learning_rate": 2.8366666666666664e-05,
-      "loss": 0.0018,
+      "loss": 0.168,
       "step": 1600
     },
     {
-      "epoch": 32.01,
+      "epoch": 32.5,
       "learning_rate": 2.815833333333333e-05,
-      "loss": 0.0018,
+      "loss": 0.1558,
       "step": 1625
     },
     {
-      "epoch": 32.01,
+      "epoch": 33.0,
       "learning_rate": 2.7949999999999995e-05,
-      "loss": 0.0017,
+      "loss": 0.1573,
       "step": 1650
     },
     {
-      "epoch": 33.01,
+      "epoch": 33.5,
       "learning_rate": 2.7741666666666662e-05,
-      "loss": 0.0017,
+      "loss": 0.1471,
       "step": 1675
     },
     {
-      "epoch": 33.01,
+      "epoch": 34.0,
       "learning_rate": 2.753333333333333e-05,
-      "loss": 0.0016,
+      "loss": 0.15,
       "step": 1700
     },
     {
-      "epoch": 34.01,
+      "epoch": 34.5,
       "learning_rate": 2.7324999999999997e-05,
-      "loss": 0.0016,
+      "loss": 0.1403,
       "step": 1725
     },
     {
-      "epoch": 34.01,
+      "epoch": 35.0,
       "learning_rate": 2.7116666666666667e-05,
-      "loss": 0.0015,
+      "loss": 0.1429,
       "step": 1750
     },
     {
-      "epoch": 35.01,
+      "epoch": 35.5,
       "learning_rate": 2.6908333333333328e-05,
-      "loss": 0.0015,
+      "loss": 0.1304,
       "step": 1775
     },
     {
-      "epoch": 35.01,
+      "epoch": 36.0,
       "learning_rate": 2.6699999999999995e-05,
-      "loss": 0.0014,
+      "loss": 0.1354,
       "step": 1800
     },
     {
-      "epoch": 36.01,
+      "epoch": 36.5,
       "learning_rate": 2.6491666666666662e-05,
-      "loss": 0.0015,
+      "loss": 0.124,
       "step": 1825
     },
     {
-      "epoch": 36.01,
+      "epoch": 37.0,
       "learning_rate": 2.628333333333333e-05,
-      "loss": 0.0013,
+      "loss": 0.1319,
       "step": 1850
     },
     {
-      "epoch": 37.01,
+      "epoch": 37.5,
       "learning_rate": 2.6075e-05,
-      "loss": 0.0014,
+      "loss": 0.1178,
       "step": 1875
     },
     {
-      "epoch": 37.01,
+      "epoch": 38.0,
       "learning_rate": 2.5866666666666667e-05,
-      "loss": 0.0013,
+      "loss": 0.1228,
       "step": 1900
     },
     {
-      "epoch": 38.01,
+      "epoch": 38.5,
       "learning_rate": 2.5658333333333328e-05,
-      "loss": 0.0013,
+      "loss": 0.1129,
       "step": 1925
     },
     {
-      "epoch": 38.01,
+      "epoch": 39.0,
       "learning_rate": 2.5449999999999995e-05,
-      "loss": 0.0012,
+      "loss": 0.116,
       "step": 1950
     },
     {
-      "epoch": 39.01,
+      "epoch": 39.5,
       "learning_rate": 2.5241666666666666e-05,
-      "loss": 0.0013,
+      "loss": 0.1077,
       "step": 1975
     },
     {
-      "epoch": 39.01,
+      "epoch": 40.0,
       "learning_rate": 2.5033333333333333e-05,
-      "loss": 0.0012,
+      "loss": 0.1103,
       "step": 2000
     },
     {
-      "epoch": 39.01,
-      "eval_loss": 0.6526896953582764,
-      "eval_runtime": 685.933,
-      "eval_samples_per_second": 9.543,
-      "eval_steps_per_second": 0.076,
-      "eval_wer": 26.71771523178808,
+      "epoch": 40.0,
+      "eval_loss": 0.5369554758071899,
+      "eval_runtime": 43.808,
+      "eval_samples_per_second": 149.425,
+      "eval_steps_per_second": 1.187,
+      "eval_wer": 27.04021799116998,
       "step": 2000
     },
     {
-      "epoch": 40.01,
+      "epoch": 40.5,
       "learning_rate": 2.4825e-05,
-      "loss": 0.0012,
+      "loss": 0.1028,
       "step": 2025
     },
     {
-      "epoch": 40.01,
+      "epoch": 41.0,
       "learning_rate": 2.4616666666666668e-05,
-      "loss": 0.0011,
+      "loss": 0.1041,
       "step": 2050
     },
     {
-      "epoch": 41.01,
+      "epoch": 41.5,
       "learning_rate": 2.4408333333333328e-05,
-      "loss": 0.0011,
+      "loss": 0.0986,
       "step": 2075
     },
     {
-      "epoch": 41.01,
+      "epoch": 42.0,
       "learning_rate": 2.42e-05,
-      "loss": 0.0011,
+      "loss": 0.1008,
       "step": 2100
     },
     {
-      "epoch": 42.01,
+      "epoch": 42.5,
       "learning_rate": 2.3991666666666666e-05,
-      "loss": 0.0011,
+      "loss": 0.0912,
       "step": 2125
     },
     {
-      "epoch": 42.01,
+      "epoch": 43.0,
       "learning_rate": 2.3783333333333333e-05,
-      "loss": 0.001,
+      "loss": 0.0969,
       "step": 2150
     },
     {
-      "epoch": 43.01,
+      "epoch": 43.5,
       "learning_rate": 2.3575e-05,
-      "loss": 0.0011,
+      "loss": 0.0892,
       "step": 2175
     },
     {
-      "epoch": 43.01,
+      "epoch": 44.0,
       "learning_rate": 2.3366666666666664e-05,
-      "loss": 0.001,
+      "loss": 0.0921,
       "step": 2200
     },
     {
-      "epoch": 44.01,
+      "epoch": 44.5,
       "learning_rate": 2.315833333333333e-05,
-      "loss": 0.001,
+      "loss": 0.0853,
       "step": 2225
     },
     {
-      "epoch": 44.01,
+      "epoch": 45.0,
       "learning_rate": 2.295e-05,
-      "loss": 0.0009,
+      "loss": 0.0865,
       "step": 2250
     },
     {
-      "epoch": 45.01,
+      "epoch": 45.5,
       "learning_rate": 2.2741666666666666e-05,
-      "loss": 0.001,
+      "loss": 0.0796,
       "step": 2275
     },
     {
-      "epoch": 45.01,
+      "epoch": 46.0,
       "learning_rate": 2.2533333333333333e-05,
-      "loss": 0.0009,
+      "loss": 0.0827,
       "step": 2300
     },
     {
-      "epoch": 46.01,
+      "epoch": 46.5,
       "learning_rate": 2.2325e-05,
-      "loss": 0.0009,
+      "loss": 0.0775,
       "step": 2325
     },
     {
-      "epoch": 46.01,
+      "epoch": 47.0,
       "learning_rate": 2.2116666666666664e-05,
-      "loss": 0.0009,
+      "loss": 0.0791,
       "step": 2350
     },
     {
-      "epoch": 47.01,
+      "epoch": 47.5,
       "learning_rate": 2.190833333333333e-05,
-      "loss": 0.0009,
+      "loss": 0.0733,
       "step": 2375
     },
     {
-      "epoch": 47.01,
+      "epoch": 48.0,
       "learning_rate": 2.17e-05,
-      "loss": 0.0009,
+      "loss": 0.0761,
       "step": 2400
     },
     {
-      "epoch": 48.01,
+      "epoch": 48.5,
       "learning_rate": 2.1491666666666666e-05,
-      "loss": 0.0009,
+      "loss": 0.0702,
       "step": 2425
     },
     {
-      "epoch": 48.01,
+      "epoch": 49.0,
       "learning_rate": 2.1283333333333333e-05,
-      "loss": 0.0008,
+      "loss": 0.0728,
       "step": 2450
     },
     {
-      "epoch": 49.01,
+      "epoch": 49.5,
       "learning_rate": 2.1075e-05,
-      "loss": 0.0008,
+      "loss": 0.0671,
       "step": 2475
     },
     {
-      "epoch": 49.01,
+      "epoch": 50.0,
       "learning_rate": 2.0866666666666665e-05,
-      "loss": 0.0008,
+      "loss": 0.0707,
       "step": 2500
     },
     {
-      "epoch": 50.01,
+      "epoch": 50.5,
       "learning_rate": 2.0658333333333332e-05,
-      "loss": 0.0008,
+      "loss": 0.0652,
       "step": 2525
     },
     {
-      "epoch": 50.01,
+      "epoch": 51.0,
       "learning_rate": 2.045e-05,
-      "loss": 0.0008,
+      "loss": 0.066,
       "step": 2550
     },
     {
-      "epoch": 51.01,
+      "epoch": 51.5,
       "learning_rate": 2.0241666666666666e-05,
-      "loss": 0.0008,
+      "loss": 0.0623,
       "step": 2575
     },
     {
-      "epoch": 51.01,
+      "epoch": 52.0,
       "learning_rate": 2.0033333333333334e-05,
-      "loss": 0.0007,
+      "loss": 0.0637,
       "step": 2600
     },
     {
-      "epoch": 52.01,
+      "epoch": 52.5,
       "learning_rate": 1.9824999999999997e-05,
-      "loss": 0.0008,
+      "loss": 0.0593,
       "step": 2625
     },
     {
-      "epoch": 52.01,
+      "epoch": 53.0,
       "learning_rate": 1.9616666666666665e-05,
-      "loss": 0.0007,
+      "loss": 0.0614,
       "step": 2650
     },
     {
-      "epoch": 53.01,
+      "epoch": 53.5,
       "learning_rate": 1.9408333333333332e-05,
-      "loss": 0.0007,
+      "loss": 0.0574,
       "step": 2675
     },
     {
-      "epoch": 53.01,
+      "epoch": 54.0,
       "learning_rate": 1.92e-05,
-      "loss": 0.0007,
+      "loss": 0.0597,
       "step": 2700
     },
     {
-      "epoch": 54.01,
+      "epoch": 54.5,
       "learning_rate": 1.8991666666666666e-05,
-      "loss": 0.0007,
+      "loss": 0.0565,
       "step": 2725
     },
     {
-      "epoch": 54.01,
+      "epoch": 55.0,
       "learning_rate": 1.8783333333333334e-05,
-      "loss": 0.0007,
+      "loss": 0.0562,
       "step": 2750
     },
     {
-      "epoch": 55.01,
+      "epoch": 55.5,
       "learning_rate": 1.8574999999999997e-05,
-      "loss": 0.0007,
+      "loss": 0.0538,
       "step": 2775
     },
     {
-      "epoch": 55.01,
+      "epoch": 56.0,
       "learning_rate": 1.8366666666666665e-05,
-      "loss": 0.0007,
+      "loss": 0.0545,
       "step": 2800
     },
     {
-      "epoch": 56.01,
+      "epoch": 56.5,
       "learning_rate": 1.8158333333333332e-05,
-      "loss": 0.0007,
+      "loss": 0.0515,
       "step": 2825
     },
     {
-      "epoch": 56.01,
+      "epoch": 57.0,
       "learning_rate": 1.795e-05,
-      "loss": 0.0006,
+      "loss": 0.0523,
       "step": 2850
     },
     {
-      "epoch": 57.01,
+      "epoch": 57.5,
       "learning_rate": 1.7741666666666663e-05,
-      "loss": 0.0007,
+      "loss": 0.0507,
       "step": 2875
     },
     {
-      "epoch": 57.01,
+      "epoch": 58.0,
       "learning_rate": 1.753333333333333e-05,
-      "loss": 0.0006,
+      "loss": 0.0513,
       "step": 2900
     },
     {
-      "epoch": 58.01,
+      "epoch": 58.5,
       "learning_rate": 1.7325e-05,
-      "loss": 0.0006,
+      "loss": 0.0482,
       "step": 2925
     },
     {
-      "epoch": 58.01,
+      "epoch": 59.0,
       "learning_rate": 1.7116666666666665e-05,
-      "loss": 0.0006,
+      "loss": 0.0487,
       "step": 2950
     },
     {
-      "epoch": 59.01,
+      "epoch": 59.5,
       "learning_rate": 1.6908333333333332e-05,
-      "loss": 0.0006,
+      "loss": 0.046,
       "step": 2975
     },
     {
-      "epoch": 59.01,
+      "epoch": 60.0,
       "learning_rate": 1.67e-05,
-      "loss": 0.0006,
+      "loss": 0.0473,
       "step": 3000
     },
     {
-      "epoch": 59.01,
-      "eval_loss": 0.6950027942657471,
-      "eval_runtime": 885.7004,
-      "eval_samples_per_second": 7.391,
-      "eval_steps_per_second": 0.059,
-      "eval_wer": 27.435154525386313,
+      "epoch": 60.0,
+      "eval_loss": 0.5768794417381287,
+      "eval_runtime": 45.1811,
+      "eval_samples_per_second": 144.884,
+      "eval_steps_per_second": 1.151,
+      "eval_wer": 26.726338300220753,
       "step": 3000
     },
     {
-      "epoch": 60.01,
+      "epoch": 60.5,
       "learning_rate": 1.6491666666666663e-05,
-      "loss": 0.0006,
+      "loss": 0.0452,
       "step": 3025
     },
     {
-      "epoch": 60.01,
+      "epoch": 61.0,
       "learning_rate": 1.6283333333333334e-05,
-      "loss": 0.0006,
+      "loss": 0.0456,
       "step": 3050
     },
     {
-      "epoch": 61.01,
+      "epoch": 61.5,
       "learning_rate": 1.6074999999999998e-05,
-      "loss": 0.0006,
+      "loss": 0.0429,
       "step": 3075
     },
     {
-      "epoch": 61.01,
+      "epoch": 62.0,
       "learning_rate": 1.5866666666666665e-05,
-      "loss": 0.0006,
+      "loss": 0.0449,
       "step": 3100
     },
     {
-      "epoch": 62.01,
+      "epoch": 62.5,
       "learning_rate": 1.5658333333333332e-05,
-      "loss": 0.0006,
+      "loss": 0.0433,
       "step": 3125
     },
     {
-      "epoch": 62.01,
+      "epoch": 63.0,
       "learning_rate": 1.545e-05,
-      "loss": 0.0006,
+      "loss": 0.0436,
       "step": 3150
     },
     {
-      "epoch": 63.01,
+      "epoch": 63.5,
       "learning_rate": 1.5241666666666665e-05,
-      "loss": 0.0006,
+      "loss": 0.0409,
       "step": 3175
     },
     {
-      "epoch": 63.01,
+      "epoch": 64.0,
       "learning_rate": 1.5033333333333332e-05,
-      "loss": 0.0005,
+      "loss": 0.042,
       "step": 3200
     },
     {
-      "epoch": 64.0,
+      "epoch": 64.5,
       "learning_rate": 1.4824999999999998e-05,
-      "loss": 0.0006,
+      "loss": 0.0404,
       "step": 3225
     },
     {
-      "epoch": 64.01,
+      "epoch": 65.0,
       "learning_rate": 1.4616666666666665e-05,
-      "loss": 0.0005,
+      "loss": 0.0409,
       "step": 3250
     },
     {
-      "epoch": 65.0,
+      "epoch": 65.5,
       "learning_rate": 1.4408333333333332e-05,
-      "loss": 0.0005,
+      "loss": 0.0392,
       "step": 3275
     },
     {
-      "epoch": 65.01,
+      "epoch": 66.0,
       "learning_rate": 1.4199999999999998e-05,
-      "loss": 0.0005,
+      "loss": 0.0394,
       "step": 3300
     },
     {
-      "epoch": 66.0,
+      "epoch": 66.5,
       "learning_rate": 1.3991666666666665e-05,
-      "loss": 0.0005,
+      "loss": 0.0381,
       "step": 3325
     },
     {
-      "epoch": 66.01,
+      "epoch": 67.0,
       "learning_rate": 1.3783333333333332e-05,
-      "loss": 0.0005,
+      "loss": 0.0383,
       "step": 3350
     },
     {
-      "epoch": 67.0,
+      "epoch": 67.5,
       "learning_rate": 1.3574999999999998e-05,
-      "loss": 0.0005,
+      "loss": 0.0375,
       "step": 3375
     },
     {
-      "epoch": 67.01,
+      "epoch": 68.0,
       "learning_rate": 1.3366666666666665e-05,
-      "loss": 0.0005,
+      "loss": 0.0373,
       "step": 3400
     },
     {
-      "epoch": 68.0,
+      "epoch": 68.5,
       "learning_rate": 1.315833333333333e-05,
-      "loss": 0.0005,
+      "loss": 0.0365,
       "step": 3425
     },
     {
-      "epoch": 68.01,
+      "epoch": 69.0,
       "learning_rate": 1.2949999999999998e-05,
-      "loss": 0.0005,
+      "loss": 0.0363,
       "step": 3450
     },
     {
-      "epoch": 69.0,
+      "epoch": 69.5,
       "learning_rate": 1.2741666666666665e-05,
-      "loss": 0.0005,
+      "loss": 0.036,
       "step": 3475
     },
     {
-      "epoch": 69.01,
+      "epoch": 70.0,
       "learning_rate": 1.253333333333333e-05,
-      "loss": 0.0005,
+      "loss": 0.0358,
       "step": 3500
     },
     {
-      "epoch": 70.0,
+      "epoch": 70.5,
       "learning_rate": 1.2324999999999998e-05,
-      "loss": 0.0005,
+      "loss": 0.0349,
       "step": 3525
     },
     {
-      "epoch": 70.01,
+      "epoch": 71.0,
       "learning_rate": 1.2116666666666667e-05,
-      "loss": 0.0005,
+      "loss": 0.0351,
       "step": 3550
     },
     {
-      "epoch": 71.0,
+      "epoch": 71.5,
       "learning_rate": 1.190833333333333e-05,
-      "loss": 0.0005,
+      "loss": 0.0332,
       "step": 3575
     },
     {
-      "epoch": 71.01,
+      "epoch": 72.0,
       "learning_rate": 1.17e-05,
-      "loss": 0.0005,
+      "loss": 0.0344,
       "step": 3600
     },
     {
-      "epoch": 72.0,
+      "epoch": 72.5,
       "learning_rate": 1.1491666666666667e-05,
-      "loss": 0.0005,
+      "loss": 0.0334,
       "step": 3625
     },
     {
-      "epoch": 72.01,
+      "epoch": 73.0,
       "learning_rate": 1.1283333333333331e-05,
-      "loss": 0.0004,
+      "loss": 0.0335,
       "step": 3650
     },
     {
-      "epoch": 73.0,
+      "epoch": 73.5,
       "learning_rate": 1.1075e-05,
-      "loss": 0.0005,
+      "loss": 0.0322,
       "step": 3675
     },
     {
-      "epoch": 73.01,
+      "epoch": 74.0,
       "learning_rate": 1.0866666666666667e-05,
-      "loss": 0.0004,
+      "loss": 0.0327,
       "step": 3700
     },
     {
-      "epoch": 74.0,
+      "epoch": 74.5,
       "learning_rate": 1.0658333333333333e-05,
-      "loss": 0.0004,
+      "loss": 0.0316,
       "step": 3725
     },
     {
-      "epoch": 74.01,
+      "epoch": 75.0,
       "learning_rate": 1.045e-05,
-      "loss": 0.0004,
+      "loss": 0.0325,
       "step": 3750
     },
     {
-      "epoch": 75.0,
+      "epoch": 75.5,
       "learning_rate": 1.0241666666666667e-05,
-      "loss": 0.0004,
+      "loss": 0.0312,
       "step": 3775
     },
     {
-      "epoch": 75.01,
+      "epoch": 76.0,
       "learning_rate": 1.0033333333333333e-05,
-      "loss": 0.0004,
+      "loss": 0.031,
       "step": 3800
     },
     {
-      "epoch": 76.0,
+      "epoch": 76.5,
       "learning_rate": 9.825e-06,
-      "loss": 0.0004,
+      "loss": 0.0307,
       "step": 3825
     },
     {
-      "epoch": 76.01,
+      "epoch": 77.0,
       "learning_rate": 9.616666666666666e-06,
-      "loss": 0.0004,
+      "loss": 0.0311,
       "step": 3850
     },
     {
-      "epoch": 77.0,
+      "epoch": 77.5,
       "learning_rate": 9.408333333333333e-06,
-      "loss": 0.0004,
+      "loss": 0.0302,
       "step": 3875
     },
     {
-      "epoch": 77.01,
+      "epoch": 78.0,
       "learning_rate": 9.199999999999998e-06,
-      "loss": 0.0004,
+      "loss": 0.0306,
       "step": 3900
     },
     {
-      "epoch": 78.0,
+      "epoch": 78.5,
       "learning_rate": 8.991666666666666e-06,
-      "loss": 0.0004,
+      "loss": 0.0299,
       "step": 3925
     },
     {
-      "epoch": 78.01,
+      "epoch": 79.0,
       "learning_rate": 8.783333333333333e-06,
-      "loss": 0.0004,
+      "loss": 0.0301,
       "step": 3950
     },
     {
-      "epoch": 79.0,
+      "epoch": 79.5,
       "learning_rate": 8.574999999999998e-06,
-      "loss": 0.0004,
+      "loss": 0.0298,
       "step": 3975
     },
     {
-      "epoch": 79.01,
+      "epoch": 80.0,
       "learning_rate": 8.366666666666666e-06,
-      "loss": 0.0004,
+      "loss": 0.03,
       "step": 4000
     },
     {
-      "epoch": 79.01,
-      "eval_loss": 0.7259679436683655,
-      "eval_runtime": 645.824,
-      "eval_samples_per_second": 10.136,
-      "eval_steps_per_second": 0.081,
-      "eval_wer": 28.40438741721854,
+      "epoch": 80.0,
+      "eval_loss": 0.5936003923416138,
+      "eval_runtime": 43.3248,
+      "eval_samples_per_second": 151.091,
+      "eval_steps_per_second": 1.2,
+      "eval_wer": 26.138245033112582,
       "step": 4000
     },
     {
-      "epoch": 80.0,
+      "epoch": 80.5,
       "learning_rate": 8.158333333333333e-06,
-      "loss": 0.0004,
+      "loss": 0.0284,
       "step": 4025
     },
     {
-      "epoch": 80.01,
+      "epoch": 81.0,
       "learning_rate": 7.949999999999998e-06,
-      "loss": 0.0004,
+      "loss": 0.0295,
       "step": 4050
     },
     {
-      "epoch": 81.0,
+      "epoch": 81.5,
       "learning_rate": 7.741666666666666e-06,
-      "loss": 0.0004,
+      "loss": 0.0279,
       "step": 4075
     },
     {
-      "epoch": 81.01,
+      "epoch": 82.0,
       "learning_rate": 7.533333333333333e-06,
-      "loss": 0.0004,
+      "loss": 0.0288,
       "step": 4100
     },
     {
-      "epoch": 82.0,
+      "epoch": 82.5,
       "learning_rate": 7.324999999999999e-06,
-      "loss": 0.0004,
+      "loss": 0.028,
       "step": 4125
     },
     {
-      "epoch": 82.01,
+      "epoch": 83.0,
       "learning_rate": 7.116666666666666e-06,
-      "loss": 0.0004,
+      "loss": 0.028,
       "step": 4150
     },
     {
-      "epoch": 83.0,
+      "epoch": 83.5,
       "learning_rate": 6.908333333333333e-06,
-      "loss": 0.0004,
+      "loss": 0.0278,
       "step": 4175
     },
     {
-      "epoch": 83.01,
+      "epoch": 84.0,
       "learning_rate": 6.699999999999999e-06,
-      "loss": 0.0004,
+      "loss": 0.0286,
       "step": 4200
     },
     {
-      "epoch": 84.0,
+      "epoch": 84.5,
       "learning_rate": 6.491666666666666e-06,
-      "loss": 0.0004,
+      "loss": 0.0275,
       "step": 4225
     },
     {
-      "epoch": 84.01,
+      "epoch": 85.0,
       "learning_rate": 6.283333333333333e-06,
-      "loss": 0.0004,
+      "loss": 0.0266,
       "step": 4250
     },
     {
-      "epoch": 85.0,
+      "epoch": 85.5,
       "learning_rate": 6.0749999999999994e-06,
-      "loss": 0.0004,
+      "loss": 0.0265,
       "step": 4275
     },
     {
-      "epoch": 85.01,
+      "epoch": 86.0,
       "learning_rate": 5.866666666666666e-06,
-      "loss": 0.0004,
+      "loss": 0.0268,
       "step": 4300
     },
     {
-      "epoch": 86.0,
+      "epoch": 86.5,
       "learning_rate": 5.658333333333332e-06,
-      "loss": 0.0004,
+      "loss": 0.026,
       "step": 4325
     },
     {
-      "epoch": 86.01,
+      "epoch": 87.0,
       "learning_rate": 5.4499999999999995e-06,
-      "loss": 0.0004,
+      "loss": 0.0265,
       "step": 4350
     },
     {
-      "epoch": 87.0,
+      "epoch": 87.5,
       "learning_rate": 5.241666666666666e-06,
-      "loss": 0.0004,
+      "loss": 0.0263,
       "step": 4375
     },
     {
-      "epoch": 87.01,
+      "epoch": 88.0,
       "learning_rate": 5.033333333333332e-06,
-      "loss": 0.0004,
+      "loss": 0.0262,
       "step": 4400
     },
     {
-      "epoch": 88.0,
+      "epoch": 88.5,
       "learning_rate": 4.825e-06,
-      "loss": 0.0004,
+      "loss": 0.0258,
       "step": 4425
     },
     {
-      "epoch": 88.01,
+      "epoch": 89.0,
       "learning_rate": 4.616666666666667e-06,
-      "loss": 0.0004,
+      "loss": 0.0255,
       "step": 4450
     },
     {
-      "epoch": 89.0,
+      "epoch": 89.5,
       "learning_rate": 4.408333333333333e-06,
-      "loss": 0.0004,
+      "loss": 0.0256,
       "step": 4475
     },
     {
-      "epoch": 89.01,
+      "epoch": 90.0,
       "learning_rate": 4.2e-06,
-      "loss": 0.0003,
+      "loss": 0.0261,
       "step": 4500
     },
     {
-      "epoch": 90.0,
+      "epoch": 90.5,
       "learning_rate": 3.991666666666666e-06,
-      "loss": 0.0004,
+      "loss": 0.0257,
       "step": 4525
     },
     {
-      "epoch": 90.01,
+      "epoch": 91.0,
       "learning_rate": 3.783333333333333e-06,
-      "loss": 0.0003,
+      "loss": 0.0256,
       "step": 4550
     },
     {
-      "epoch": 91.0,
+      "epoch": 91.5,
       "learning_rate": 3.575e-06,
-      "loss": 0.0004,
+      "loss": 0.0253,
       "step": 4575
     },
     {
-      "epoch": 91.01,
+      "epoch": 92.0,
       "learning_rate": 3.3666666666666665e-06,
-      "loss": 0.0003,
+      "loss": 0.0249,
       "step": 4600
     },
     {
-      "epoch": 92.0,
+      "epoch": 92.5,
       "learning_rate": 3.1583333333333333e-06,
-      "loss": 0.0004,
+      "loss": 0.0252,
       "step": 4625
     },
     {
-      "epoch": 92.01,
+      "epoch": 93.0,
       "learning_rate": 2.9499999999999997e-06,
-      "loss": 0.0003,
+      "loss": 0.0254,
       "step": 4650
     },
     {
-      "epoch": 93.0,
+      "epoch": 93.5,
       "learning_rate": 2.7416666666666665e-06,
-      "loss": 0.0003,
+      "loss": 0.0249,
       "step": 4675
     },
     {
-      "epoch": 93.01,
+      "epoch": 94.0,
       "learning_rate": 2.533333333333333e-06,
-      "loss": 0.0003,
+      "loss": 0.0247,
       "step": 4700
     },
     {
-      "epoch": 94.0,
+      "epoch": 94.5,
       "learning_rate": 2.3249999999999998e-06,
-      "loss": 0.0003,
+      "loss": 0.0252,
       "step": 4725
     },
     {
-      "epoch": 94.01,
+      "epoch": 95.0,
       "learning_rate": 2.1166666666666666e-06,
-      "loss": 0.0003,
+      "loss": 0.0247,
       "step": 4750
     },
     {
-      "epoch": 95.0,
+      "epoch": 95.5,
       "learning_rate": 1.908333333333333e-06,
-      "loss": 0.0003,
+      "loss": 0.0247,
       "step": 4775
     },
     {
-      "epoch": 95.01,
+      "epoch": 96.0,
       "learning_rate": 1.7e-06,
-      "loss": 0.0003,
+      "loss": 0.0246,
       "step": 4800
     },
     {
-      "epoch": 96.0,
+      "epoch": 96.5,
       "learning_rate": 1.4916666666666666e-06,
-      "loss": 0.0003,
+      "loss": 0.0244,
       "step": 4825
     },
     {
-      "epoch": 96.01,
+      "epoch": 97.0,
       "learning_rate": 1.2833333333333333e-06,
-      "loss": 0.0003,
+      "loss": 0.0236,
       "step": 4850
     },
     {
-      "epoch": 97.0,
+      "epoch": 97.5,
       "learning_rate": 1.0749999999999999e-06,
-      "loss": 0.0003,
+      "loss": 0.0239,
       "step": 4875
     },
     {
-      "epoch": 97.01,
+      "epoch": 98.0,
       "learning_rate": 8.666666666666666e-07,
-      "loss": 0.0003,
+      "loss": 0.0238,
       "step": 4900
     },
     {
-      "epoch": 98.0,
+      "epoch": 98.5,
       "learning_rate": 6.583333333333333e-07,
-      "loss": 0.0003,
+      "loss": 0.024,
       "step": 4925
     },
     {
-      "epoch": 98.01,
+      "epoch": 99.0,
       "learning_rate": 4.5e-07,
-      "loss": 0.0003,
+      "loss": 0.0246,
       "step": 4950
     },
     {
-      "epoch": 99.0,
+      "epoch": 99.5,
       "learning_rate": 2.4166666666666665e-07,
-      "loss": 0.0003,
+      "loss": 0.0242,
       "step": 4975
     },
     {
-      "epoch": 99.01,
+      "epoch": 100.0,
       "learning_rate": 3.3333333333333334e-08,
-      "loss": 0.0003,
+      "loss": 0.0244,
       "step": 5000
     },
     {
-      "epoch": 99.01,
-      "eval_loss": 0.7315477132797241,
-      "eval_runtime": 782.825,
-      "eval_samples_per_second": 8.362,
-      "eval_steps_per_second": 0.066,
-      "eval_wer": 28.1905353200883,
+      "epoch": 100.0,
+      "eval_loss": 0.6002562642097473,
+      "eval_runtime": 43.2619,
+      "eval_samples_per_second": 151.311,
+      "eval_steps_per_second": 1.202,
+      "eval_wer": 26.13307119205298,
       "step": 5000
     },
     {
-      "epoch": 99.01,
+      "epoch": 100.0,
       "step": 5000,
       "total_flos": 3.1497401622528e+19,
-      "train_loss": 0.05570581116452813,
-      "train_runtime": 137346.4264,
-      "train_samples_per_second": 9.319,
-      "train_steps_per_second": 0.036
+      "train_loss": 0.5713151776909828,
+      "train_runtime": 5322.8653,
+      "train_samples_per_second": 240.472,
+      "train_steps_per_second": 0.939
     }
   ],
   "logging_steps": 25,
   "max_steps": 5000,
-  "num_train_epochs": 9223372036854775807,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 100,
   "save_steps": 1000,
   "total_flos": 3.1497401622528e+19,
+  "train_batch_size": 256,
   "trial_name": null,
   "trial_params": null
 }